A product recommendation method and device

Abstract

The present invention discloses a commodity recommendation method and device, and relates to the field of computer technology. A specific implementation of the method includes: determining the candidate category of the first user's next purchase of the commodity based on the category transfer data of the past purchase of the commodity by each second user in the first user and its nearest neighbor set and the set recommendation time, and determining the candidate brand of the first user's next purchase of the commodity in the candidate category based on the vector similarity between the first brand and each brand in the brand purchase network; calculating the purchase probability of the first user for each target commodity of the candidate brand based on the price matching degree of the first user for each target commodity of the candidate brand, and the vector similarity between the first brand and the candidate brand, so as to recommend the commodity based on the purchase probability. This implementation can capture the changes in user interests between different categories, take the purchase time interval into consideration, fully explore the user's price preference, and provide diverse, rich, and accurate recommendation results based on the changes in user interests.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a commodity recommendation method and device. Background Art

[0002] The purpose of product recommendation is to explore user preferences and recommend products that they are interested in. The current product recommendation strategy is mainly based on user behavior information to explore user preferences.

[0003] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art:

[0004] The lack of analytical research on category conversion and purchase time intervals, as well as the lack of in-depth research on product prices, makes the product recommendation results too simple and repetitive, lacking diversity, and the recommendation results are not accurate enough. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a product recommendation method and device, which can capture the changes in user interests between different categories, take the purchase time interval into consideration, fully explore the user's price preferences, and provide diverse, rich and accurate recommendation results based on the user's interest changes.

[0006] To achieve the above objective, according to one aspect of an embodiment of the present invention, a product recommendation method is provided.

[0007] A product recommendation method includes: determining a candidate category for the first user to purchase next time based on category transfer data of products purchased in the past by each second user in a first user and its nearest neighbor set and a set recommendation time; determining a candidate brand for the first user to purchase products of the candidate category next time by the first user using vector similarities between a first brand and each brand in a brand purchase network, wherein the brand purchase network is constructed based on all user purchase records, and the first brand is the brand that the first user purchased most recently in the past; and calculating the purchase probability of the first user for each target product of the candidate brand based on the price matching degree of the first user for each target product of the candidate brand and the vector similarity between the first brand and the candidate brand in the brand purchase network, so as to recommend the products of the candidate brand to the first user according to the purchase probability.

[0008] Optionally, the category transfer data of the user's past purchases of goods includes the category transfer pairs in the user's purchase category sequence, the frequency of occurrence of the category transfer pairs in the purchase category sequence, and the purchase time of the two categories in the category transfer pairs by the user, and the users include the first user and the second user.

[0009] Optionally, the determining the candidate category of the next purchase of the first user according to the category transfer data of the past purchases of the goods of each second user in the first user and its nearest neighbor set and the set recommendation time includes: for each second user, obtaining a first set from the category transfer pairs in the purchase category sequence of the first user, and obtaining a second set from the category transfer pairs in the purchase category sequence of the second user; calculating the similarity of the purchase category sequences of the first user and the second user according to the sum of the occurrence frequencies in the intersection of the first set and the second set, and the sum of the occurrence frequencies in the first set and the second set; for each second user, calculating the purchase time interval corresponding to the target category transfer pair including the first category in the purchase category sequence of the second user, and the target category transfer pair in the purchase category sequence of the second user. According to the occurrence frequency of the first user in the purchase category sequence, the initial purchase time interval preference of the second user on the target category transfer pair is calculated, where the first category is the category where the first brand is located, and the purchase time interval corresponding to the target category transfer pair is the difference between the purchase time of the second user in the two categories in the target category transfer pair; according to the similarity of the purchase category sequence of the first user and each of the second users, and the initial purchase time interval preference of each of the second users on the target category transfer pair, the final purchase time interval preference of the first user on the target category transfer pair is calculated; according to the purchase time of the first user in the first category, the set recommended time, and the final purchase time interval preference of the first user on the target category transfer pair, the probability of each category of the first user's next purchase of goods is calculated, and the candidate category is selected from the various categories according to the probability.

[0010] Optionally, the method of utilizing the vector similarity between the first brand and each brand in the brand purchase network to determine the candidate brand of the product of the candidate category that the first user will purchase next time includes: learning the constructed brand purchase network through a preset model to obtain a brand vector representation with category personalization; and calculating the vector similarity between the first brand and each brand using the brand vector representation with category personalization.

[0011] Optionally, it also includes calculating the price matching degree of the first user for each target product of the candidate brand in the following manner: for each target product, calculating the degree of membership of the target product's price to each price level based on the price of the target product and the lowest price and the highest price of products in the category to which the target product belongs; calculating the user price sensitivity of the first user corresponding to each price level based on the degree of membership of the price of each target product to each price level and the first user's satisfaction with each target product; calculating the price matching degree of the first user for each target product of the candidate brand based on the user price sensitivity of the first user corresponding to each price level and the degree of membership of the price of each target product to each price level.

[0012] Optionally, the calculating the purchase probability of the first user for each target product of the candidate brand based on the price matching degree of the first user for each target product of the candidate brand and the vector similarity between the first brand and the candidate brand in the brand purchasing network includes: for each target product, summing and weighted summing the price matching degree of the first user for the target product and the vector similarity between the first brand in the brand purchasing network and the brand to which the target product belongs; and calculating the purchase probability of the first user for each target product according to the result of the weighted sum and the result of the sum.

[0013] According to another aspect of an embodiment of the present invention, a commodity recommendation device is provided.

[0014] A product recommendation device includes: a category prediction module, which is used to determine the candidate category of the product that the first user will purchase next time based on the category transfer data of the products purchased in the past by each second user in the first user and its nearest neighbor set and the set recommendation time; a brand prediction module, which is used to determine the candidate brand of the product of the candidate category that the first user will purchase next time by using the vector similarity between a first brand and each brand in a brand purchase network, wherein the brand purchase network is constructed based on the purchase records of all users, and the first brand is the brand of the product that the first user purchased most recently in the past; a product recommendation module, which is used to calculate the purchase probability of the first user for each target product of the candidate brand based on the price matching degree of the first user for each target product of the candidate brand and the vector similarity between the first brand and the candidate brand in the brand purchase network, so as to recommend the products of the candidate brand to the first user according to the purchase probability.

[0015] Optionally, the category transfer data of the user's past purchases of goods includes the category transfer pairs in the user's purchase category sequence, the frequency of occurrence of the category transfer pairs in the purchase category sequence, and the purchase time of the two categories in the category transfer pairs by the user, and the users include the first user and the second user.

[0016] Optionally, the category prediction module is further used to: for each second user, obtain a first set from the category transfer pairs in the purchase category sequence of the first user, obtain a second set from the category transfer pairs in the purchase category sequence of the second user, and calculate the similarity of the purchase category sequences of the first user and the second user according to the sum of the occurrence frequencies in the intersection of the first set and the second set, and the sum of the occurrence frequencies in the first set and the second set; for each second user, calculate the purchase time interval corresponding to the target category transfer pair of the first category in the purchase category sequence of the second user, and the occurrence frequency of the target category transfer pair in the purchase category sequence of the second user, an initial purchase time interval preference on a target category transfer pair, wherein the first category is the category where the first brand is located, and the purchase time interval corresponding to the target category transfer pair is the difference between the purchase times of the second user in the two categories in the target category transfer pair; according to the similarity of the purchase category sequences of the first user and each of the second users, and the initial purchase time interval preference of each of the second users on the target category transfer pair, calculating the final purchase time interval preference of the first user on the target category transfer pair; according to the purchase time of the first user in the first category, the set recommended time, and the final purchase time interval preference of the first user on the target category transfer pair, calculating the probability of each category that the first user will purchase the product next time, and selecting the candidate category from the various categories according to the probability.

[0017] Optionally, the brand prediction module is further used to: learn the constructed brand purchase network through a preset model to obtain a brand vector representation with category personalization; and use the brand vector representation with category personalization to calculate the vector similarity between the first brand and each brand.

[0018] Optionally, it also includes a price matching calculation module, which is used to calculate the price matching degree of the first user for each target product of the candidate brand in the following manner: for each target product, calculate the degree of membership of the target product's price to each price level based on the price of the target product and the lowest price and the highest price of products in the category to which the target product belongs; calculate the user price sensitivity of the first user corresponding to each price level based on the degree of membership of the price of each target product to each price level and the first user's satisfaction with each target product; calculate the price matching degree of the first user for each target product of the candidate brand based on the user price sensitivity of the first user corresponding to each price level and the degree of membership of the price of each target product to each price level.

[0019] Optionally, the product recommendation module is also used to: for each of the target products, sum and weighted sum the price matching degree of the first user for the target product and the vector similarity between the first brand in the brand purchase network and the brand to which the target product belongs; and calculate the purchase probability of the first user for each of the target products based on the result of the weighted sum and the result of the sum.

[0020] According to yet another aspect of the embodiments of the present invention, an electronic device is provided.

[0021] An electronic device includes: one or more processors; a memory for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the commodity recommendation method provided by an embodiment of the present invention.

[0022] According to yet another aspect of an embodiment of the present invention, a computer readable medium is provided.

[0023] A computer-readable medium stores a computer program, which, when executed by a processor, implements the commodity recommendation method provided in an embodiment of the present invention.

[0024] An embodiment of the above invention has the following advantages or beneficial effects: according to the category transfer data of the past purchases of the goods by each second user in the first user and its nearest neighbor set and the set recommendation time, the candidate category of the goods that the first user will purchase next time is determined, and the candidate brand of the goods of the candidate category that the first user will purchase next time is determined by using the vector similarity between the first brand and each brand in the brand purchase network; based on the price matching degree of each target product of the candidate brand by the first user and the vector similarity between the first brand and the candidate brand, the purchase probability of each target product by the first user is calculated to recommend the goods according to the purchase probability. It can capture the changes in user interests between different categories, take the purchase time interval into consideration, fully explore the price preference of the user, and provide diverse, rich and accurate recommendation results according to the changes in user interests.

[0025] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with specific implementation examples. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0027] Figure 1 is a schematic diagram of the main steps of a commodity recommendation method according to an embodiment of the present invention;

[0028] Figure 2 is a schematic diagram of a category prediction process according to an embodiment of the present invention;

[0029] Figure 3 is a schematic diagram of a brand prediction process according to an embodiment of the present invention;

[0030] Figure 4 is a schematic diagram of a price matching prediction process between a user and a product according to an embodiment of the present invention;

[0031] Figure 5 is a schematic diagram of a product recommendation process according to an embodiment of the present invention;

[0032] Figure 6 is a schematic diagram of the overall architecture of product recommendation according to an embodiment of the present invention;

[0033] Figure 7 is a schematic diagram of main modules of a commodity recommendation device according to an embodiment of the present invention;

[0034] Figure 8 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0035] Fig. 9It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0037] Figure 1 1 is a schematic diagram of main steps of a commodity recommendation method according to an embodiment of the present invention.

[0038] like Figure 1 As shown, a commodity recommendation method according to an embodiment of the present invention mainly includes the following steps S101 to S103.

[0039] Step S101: Determine candidate categories for the next purchase of goods by the first user based on the category transfer data of the first user and each second user in the nearest neighbor set in the past and the set recommendation time.

[0040] Step S102: using the vector similarity between the first brand and each brand in the brand purchase network, determine a candidate brand for the first user to purchase the product of the candidate category next time.

[0041] Step S103: Based on the price matching degree of the first user for each target product of the candidate brand and the vector similarity between the first brand and the candidate brand in the brand purchase network, the purchase probability of the first user for each target product is calculated to recommend the products of the candidate brand to the first user according to the purchase probability.

[0042] The first user is the user of the current product to be recommended.

[0043] The second user is a user in the nearest neighbor set of the first user. The nearest neighbor set may be determined by a nearest neighbor algorithm, such as a K-nearest neighbor classification algorithm.

[0044] The brand purchase network is constructed based on all user purchase records, and the first brand is the brand of the first user's most recent purchase in the past.

[0045] The category transfer data of the user's past purchases of goods include the category transfer pairs in the user's purchase category sequence, the frequency of occurrence of the category transfer pairs in the purchase category sequence, and the purchase time of the two categories in the category transfer pairs by the user. The user includes the first user and the second user. The category transfer data of the first user's past purchases of goods include the category transfer pairs in the first user's purchase category sequence, the frequency of occurrence of the category transfer pairs in the purchase category sequence, and the purchase time of the two categories in the category transfer pairs by the first user. The category transfer data of the second user's past purchases of goods is similar.

[0046] In one embodiment, based on the category transfer data of the past purchases of goods by each second user in the first user and its nearest neighbor set and the set recommendation time, the candidate category of the next purchase of goods by the first user is determined, specifically including: for each second user, a first set is obtained from the category transfer pairs in the purchase category sequence of the first user, and a second set is obtained from the category transfer pairs in the purchase category sequence of the second user; based on the sum of the above occurrence frequencies in the intersection of the first set and the second set, and the sum of the above occurrence frequencies in the first set and the second set, the similarity of the purchase category sequences of the first user and the second user is calculated, specifically see S202 below; for each second user, based on the purchase time interval corresponding to the target category transfer pairs including the first category in the purchase category sequence of the second user, the similarity of the purchase category sequences of the first user and the second user is calculated, and the frequency of occurrence of the target category transfer pair in the purchase category sequence of the second user, calculate the initial purchase time interval preference of the second user on the target category transfer pair, see S201 below for details; calculate the final purchase time interval preference of the first user on the target category transfer pair according to the similarity of the purchase category sequences between the first user and each second user, and the initial purchase time interval preference of each second user on the target category transfer pair, see S203 below for details; calculate the probability of each category of the first user's next purchase of goods according to the purchase time of the first user in the first category, the set recommendation time, and the final purchase time interval preference of the first user on the target category transfer pair, and select candidate categories from the various categories of the first user's next purchase of goods according to the probability, see S204 and S205 below for details.

[0047] The first category is the category where the first brand is located, and the purchase time interval corresponding to the target category transfer pair is the difference between the purchase times of the second user in the two categories in the target category transfer pair.

[0048] In other embodiments, the steps of calculating the purchase category sequence similarity and the initial purchase time interval preference may be interchanged.

[0049] In one embodiment, the vector similarity between the first brand and each brand in the brand purchase network is used to determine the candidate brand of the candidate category that the first user will purchase next time. Specifically, the following steps are performed: learning the constructed brand purchase network through a preset model to obtain a brand vector representation with category personalization; and calculating the vector similarity between the first brand and each brand using the brand vector representation with category personalization.

[0050] The preset model can be implemented based on the LINE (Large-scale Information Network Embedding) model that retains the second-order similarity.

[0051] In one embodiment, the price matching degree of the first user for each target product of the candidate brand may be calculated in the following manner: for each target product, the degree of membership of the target product's price to each price level is calculated based on the price of the target product and the lowest price and the highest price of products in the category to which the target product belongs; the user price sensitivity of the first user corresponding to each price level is calculated based on the degree of membership of the price of each target product to each price level and the first user's satisfaction with each target product; the price matching degree of the first user for each target product of the candidate brand is calculated based on the user price sensitivity of the first user corresponding to each price level and the degree of membership of the price of each target product to each price level.

[0052] In one embodiment, based on the price matching of the first user for each target product of the candidate brand and the vector similarity between the first brand and the candidate brand in the brand purchase network, the purchase probability of the first user for each target product is calculated, specifically including: for each target product, summing and weighting the price matching of the first user for the target product and the vector similarity between the first brand in the brand purchase network and the brand to which the target product belongs; and calculating the purchase probability of the first user for each target product according to the result of the weighted sum and the result of the sum. For the specific calculation method, see S504.

[0053] The embodiments of the present invention can overcome the homogeneity defect of recommendation results existing in the prior art, capture the changes in user interests between different categories, and provide diverse, rich and accurate recommendation results based on the changes in user interests.

[0054] Figure 2 4 is a schematic diagram of a category prediction process according to an embodiment of the present invention.

[0055] Through the category prediction process of an embodiment of the present invention, the candidate category of the commodity that the user will purchase next time can be predicted.

[0056] The user's interest in different categories of goods changes with time, that is, at different time points, the user's interest in the same type of goods will also show different preferences. Therefore, the embodiment of the present invention combines this time preference to establish the user's interest transfer relationship in categories at different time intervals.

[0057] like Figure 2 As shown, the category prediction process of the embodiment of the present invention includes:

[0058] S201: Calculate the user's initial purchase time interval preference on the category transfer pair.

[0059] The embodiment of the present invention mines the purchase time interval preference of a single user in category transfer according to the user's purchase category sequence as the initial purchase time interval preference of the user in category transfer.

[0060] By scanning the purchase category sequence of user u (u represents any user), the category transfer pairs therein, the occurrence frequency of each category transfer pair in the purchase category sequence, and the purchase time of the user in each category can be obtained.

[0061] A category transfer pair refers to a category pair consisting of two adjacent categories in a purchase category sequence. Since the categories in the purchase category sequence are arranged in the order in which the user purchases, the category pair consisting of two adjacent categories reflects the category transfer relationship when the user purchases the product. Taking the purchase category sequence {a, b, b, a, c, d, a, b} as an example, one of the category transfer pairs can be recorded as follows: a->b, or {a, b}, that is, the user first purchases a product in category a, and then purchases a product in category b, and a and b form a category transfer pair. The same is true for other category transfer pairs, which are not listed one by one. Since the category transfer relationship a->b appears twice in the above purchase category sequence, the frequency of occurrence of a->b in the purchase category sequence is 2.

[0062] The initial purchase time interval preference of user u on the category transfer pair a->b is calculated by the following formula:

[0063]

[0064] Where a and b represent two adjacent categories in the purchase category sequence of user u, a comes first and b comes later (i.e. user u purchases goods in category a first and then in category b), and the category transfer pair is a->b. N(b,a) represents the frequency of a->b in the purchase category sequence of user u. T a and T b represent the purchase time (i.e. the time of purchasing goods) of user u in category a and category b respectively, T b -T aIt represents the purchase time interval corresponding to the category transfer pair a->b, that is, the length of time between each time when user u purchases goods in category a and then in category b in the purchase category sequence. It represents the initial purchase time interval preference of user u on the category transfer pair a->b.

[0065] Similarly, the initial purchase time interval preference of user u in other category transfer pairs can be calculated.

[0066] S202: Calculate the purchase category sequence similarity between users.

[0067] The similarity of users in their purchase category sequences indicates the similarity between users in terms of category interest transfer.

[0068] Category transfer focuses on the transfer relationship between two categories of users. The embodiment of the present invention decomposes the user's purchase category sequence into multiple category transfer pairs. Category transfer pairs can also include their frequency of occurrence information in the purchase category sequence. For example, if the purchase category sequence of a user u1 is {a,b,c,d}, it can be decomposed into {a,b,n1},{b,c,n2},{c,d,n3}, where n1, n2, and n3 represent the frequency of occurrence of {a,b}, {b,c}, and {c,d} in the purchase category sequence, respectively. The purchase category sequence of each user is decomposed as above, and then the similarity between users is measured on the decomposed sequence. Assume that the intersection of the purchase category sequences of two users u1 and u2 is S c , which includes the intersection of the category transfer pairs of the two users (excluding the frequency of occurrence), namely: [{a,b}, {b,c}, {d,e}]. Then the purchase category sequence similarity between the two users is as follows:

[0069]

[0070] Among them, the numerator part i represents each category transfer pair in the intersection, and nu 1i and nu 2i They represent the occurrence frequencies of category transfer pair i in the purchase category sequence of user u1 and in the purchase category sequence of user u2 respectively. Therefore, the numerator represents the sum of the occurrence frequencies of the category transfer pairs in the above intersection of the two users, while the denominator represents the sum of the occurrence frequencies of all category transfer pairs in the purchase category sequences of the two users.

[0071] For example, assuming that the category transfer pairs after user u1 is decomposed form the first set: {a, b, n1}; {b, c, n2}; {c, d, n3}, and the category transfer pairs after user u2 is decomposed form the second set: {a, b, m1}; {b, c, m2}; {r, t, m3}, we can get the intersection Sc =[{a,b},{b,c}], then the purchase category sequence similarity between the two users is:

[0072]

[0073] S203: Calculate the user's final purchase time interval preference on the category transfer pair.

[0074] Since each user has limited behavior, it is impossible to cover all category transfer situations, which is especially obvious for new users. Therefore, the initial purchase time interval preference is incomplete. The purchase time interval preference is expanded based on the initial purchase time interval preference and the similarity of the purchase category sequences between users, so as to obtain the user's final purchase time interval preference.

[0075] Based on the initial purchase interval preference above Similarity between purchase categories and user sequences The following calculation formula for the final purchase time interval preference can be obtained:

[0076]

[0077] Among them, a and b represent the two categories in the category transfer pair a->b (the order is determined according to the user's purchase time). represents the k nearest neighbor set of user u1, represents the initial purchase time interval preference of user u2 on the category transfer pair a->b, For the specific calculation method of , see S201. represents the purchase category sequence similarity between two users, For the calculation method of , see S202. Represents the final purchase time interval preference of user u1 on the category transfer pair a->b.

[0078] S204: Calculate the probability of the user purchasing each category of goods next time.

[0079] Based on the user’s last purchased category and corresponding purchase time, as well as the set recommended time T b ', the user's final purchase time interval preference on the category transfer pair can be used to calculate the probability of the category of the product that the user will purchase next. For example, the probability of user u1 purchasing the product of category b next is calculated by taking the category a purchased by user u1 as an example. The probability calculation formula is as follows:

[0080]

[0081] Similarly, we can calculate the probability that user u1 will purchase products in other categories next.

[0082] S205: sorting the probabilities of various categories of goods that the user will purchase next time, and selecting a first preset number of categories with the largest probabilities as candidate categories for the user to purchase next time.

[0083] For example, based on the probability of user u1 purchasing goods in each category next, the 10 categories with the highest probability can be selected as candidate categories for user u1 to purchase goods next.

[0084] Figure 3 4 is a schematic diagram of a brand prediction process according to an embodiment of the present invention.

[0085] Through the brand prediction process of this embodiment, the candidate brands of the commodities under the candidate categories that the user will purchase next time can be determined.

[0086] S301: Build a brand purchase network and learn a brand vector representation with category personalization by retaining the LINE model of second-order similarity.

[0087] Users usually buy several brands in a certain category. A large number of user purchase records can form a huge brand purchase network, which contains the potential associations between the brands purchased by users. This association is crucial for exploring users' potential interests. To build a brand purchase network, a weighted directed graph of brands can be constructed based on the purchase order of users, and the potential associations can be learned through a preset algorithm.

[0088] The weight in the brand purchase network of the embodiment of the present invention is defined as the ratio of the number of ordering users. Assume that v i and v j Represents two nodes in a weighted directed graph, e i,j represents the edge connecting these two nodes, then the weight of the edge (network weight) can be expressed as:

[0089]

[0090] Among them, Order(v i →v j ) means buying brand v i After purchasing brand v j The number of users, and Order(v i ) means buying brand v i The number of users.

[0091] Through the above steps, a directed brand purchase network (weighted directed graph) with personalized weights is constructed. The embodiment of the present invention can use the graph embedding method to learn the potential associations in the weighted directed graph, and finally learn a set of vector representations about brands. The brand vector representation can express the potential associations between brands. Graph embedding obtains the vector representation of each node by learning the node relationship in the network. And the distance between vectors represents the relationship between nodes.

[0092] In the graph embedding architecture, the embodiment of the present invention can be specifically based on the LINE model (hereinafter referred to as the model) that retains the second-order similarity. The model can efficiently learn the weighted directed graph. In the model, for each vertex (i.e., brand), there are two vectors h corresponding to each vertex. i and h i ', respectively represent v i The vector representation when it is a vertex and the vector representation when it is a neighbor node. Finally, the model outputs the vector representation of each vertex.

[0093] In the scenario of brand purchase, the embodiment of the present invention distinguishes different categories, that is, there is a certain gap between the brand purchase networks corresponding to different categories, and the definition of the above network weights also depends on the category. Therefore, different categories correspond to different brand purchase networks, with different topological structures and weights, but the network nodes are basically the same (nodes represent brands). However, learning a model for each category is often not feasible in practical applications, which wastes resources and increases the complexity of the model. In order to solve this problem, the embodiment of the present invention adopts the method of sharing vertices and vectors to synchronously learn the brand purchase networks of all categories.

[0094] The network weights and topologies are different in different categories, but the vertex vector h i and h i 'Use the sharing method and add category vectors to personalize different categories. i For example, suppose the vector h i The dimension of is d, and the brand vector after category personalization is calculated as follows:

[0095]

[0096]

[0097] Where M represents the total number of categories, d represents the dimension of the category matrix W, and is also the final result vector hc i Each row in W represents a category vector, which indicates the embedding representation of the category. kIt is an M-dimensional one-hot vector (unique hot encoding vector) that indicates the category to which the vertex belongs.

[0098] The improvement of the loss function of the LINE model in the embodiment of the present invention is as follows:

[0099]

[0100] Among them, the first term Σ represents the positive sample, that is, the connection in the brand purchase network, the second term Σ represents the negative sample (negative sampling is performed in each category), and σ is the sigmoid function. In the learning process, the iterative algorithm can use the batch gradient descent method to finally output the optimized category personalized brand vector hc i and hc i ', hc i Indicates v i As the brand vector of the personalized category at the vertex, hc i ' indicates v i The personalized brand vector of the category when it is a neighbor node.

[0101] The embodiment of the present invention incorporates the influence of the category into the vector learning of the model to obtain a new vertex vector hc i , so that the brand vector representation under different categories can be learned in a personalized way. Although each category corresponds to a different brand purchase network, they share the same set of vertex vectors, so that the personalized impact of the category can be learned without greatly increasing the complexity of the model. As for other aspects of the model, such as negative sampling, they are treated equally for the brand purchase networks of each category and can be carried out in parallel. The category matrix W participates in the model update, and the hc obtained after the model is proficient i It is a brand vector representation with category personalization.

[0102] S302: Calculate the vector similarity between brands using the brand vector representation with category personalization.

[0103] The above has obtained a brand vector representation with category personalization, so the similarity between brand vectors (i.e., vector similarity) represents the purchase relationship between brands, that is, the probability that a user will buy another brand after buying one brand. i With brand v j The vector similarity between them, that is, the brand the user purchased last time was v i , then the user will buy brand v next j The probability is:

[0104]

[0105] S303: Determine a candidate brand of a product in the candidate category that the user will purchase next time by using the vector similarity between the brand of the product that the user purchased most recently and the brands under the candidate category that the user will purchase next time.

[0106] For example, user u1's next candidate category for purchasing goods is one of the 10 categories, and the brand he purchased last time is v i , through the brand prediction method of the embodiment of the present invention, v i The vector similarity with each brand under the 10 categories, that is, the probability of the user purchasing each brand under the 10 categories next, is arranged from large to small by the probability value, and the 10 brands with the largest probability are taken as the candidate brands for the user to purchase the goods next time. These candidate brands are all brands under the candidate categories.

[0107] Figure 4 The figure is a schematic diagram of a price matching prediction process between a user and a product according to an embodiment of the present invention.

[0108] Through the price matching prediction process between users and commodities in this embodiment, the price matching degree of the user for each commodity of the candidate brand can be obtained.

[0109] The shopping decisions of users are largely affected by the prices of goods, but since it is difficult to model prices, common recommendation systems often ignore the impact of prices. The embodiments of the present invention mine the impact of price factors on user decisions and can mine the price sensitivity of users to goods.

[0110] S401: Calculate the degree of membership of the commodity price to each price level.

[0111] As a continuous value, price is difficult to measure with a unified indicator. Therefore, the embodiment of the present invention uses fuzzy theory to represent it. First, the price level is divided into several price levels, and then the degree to which the commodity belongs to each price level, that is, the degree of membership, is calculated. The number of price levels can be adjusted according to the scene and effect. For the sake of convenience, three price levels are used as an example for introduction.

[0112] The price level is divided into three price levels from high to low: L1, L2, and L3. Each price level corresponds to a fuzzy set. Since the price levels of different categories vary greatly, different price fuzzy sets can be established according to different categories. Then, the fuzzy set of price level "L1" of category C can be characterized by the following formula:

[0113]

[0114] Among them, p is a random variable, representing the price of the commodity, and the membership degree L 1C(p) indicates the degree to which price p belongs to the L1 set, that is, the degree to which the price belongs to the "L1" price level. In addition, P L and P H Respectively represent the lowest price and highest price of all products in category C.

[0115] Using a similar method, the fuzzy set of the "L2" level for category C can be obtained as follows:

[0116]

[0117]

[0118] Among them, the membership degree L 2C (p) indicates the degree to which price p belongs to the "L2" level.

[0119] Similarly, the fuzzy set for the "L3" level of category C is as follows:

[0120]

[0121] Among them, the membership degree L 3C (p) indicates the degree to which price p belongs to the "L3" level.

[0122] After the above process, the commodity can be represented according to its price p, that is, the price level of the commodity in category C can be characterized by the following 3D vector:

[0123] Emb c (p)=(L 1C (p),L 2C (p),L 3C (p))

[0124] If more price levels are used, the corresponding vector dimensions will increase accordingly.

[0125] S402: Calculate the user price sensitivity corresponding to each price level of the user.

[0126] User price sensitivity (or price preference) can be obtained based on the price of the goods they purchase:

[0127]

[0128]

[0129]

[0130] UEmb c =(UL 1C ,UL 2C ,UL 3C );

[0131] Among them, I C Represents the set of products purchased by the user in the category, r i represents the user's satisfaction with product i, L 1C (p i )、L 2C (p i )、L 3C (p i ) respectively represent the degree of membership of the price of commodity i to each price level. In the embodiment of the present invention, the satisfaction level can be obtained by reading the user's rating of the commodity, or by other display feedback methods such as the user's text evaluation of the commodity. Then the user's price preference can be obtained through the above weighted aggregation according to the price of the commodity purchased by the user.

[0132] S403: Calculate the price matching degree between the user and the product.

[0133] For a given product P, assuming its price is p and its category is C, then the price matching degree of user u for the product is:

[0134]

[0135] For example, through category prediction, the candidate categories for user u1 to purchase next are 10 categories, and the brand purchased last time is v i Through brand prediction, 10 candidate brands of the products under the 10 candidate categories that the user will purchase next time are obtained. Through the price matching prediction of users and products in this embodiment, the price matching degree between user u1 and each product under the 10 candidate brands can be obtained.

[0136] Figure 5 The figure is a schematic diagram of a product recommendation process according to an embodiment of the present invention.

[0137] Through the commodity recommendation process of this embodiment, the purchase probability of a user for a commodity can be calculated, and then the commodity recommended to the user can be determined.

[0138] S501: Calculate in real time the probabilities of various categories corresponding to the products that the user will purchase next, and select KC categories with the largest probabilities as candidate categories.

[0139] S502: Predict the probability of the user purchasing each brand in the candidate category next, and select KB brands with the largest probability as candidate brands.

[0140] The above KC and KB are user-defined values. After the above two recall steps, the range of candidate products is greatly reduced, and the candidate products are the products of the candidate brands (ie, target products).

[0141] S503: For the candidate commodities, calculate the price matching degree of the user for each commodity.

[0142] S504: Based on the price matching degree of the user for the target product and the probability that the user will purchase the brand of the target product next, a comprehensive weighting is performed to obtain the purchase probability of the user for each target product.

[0143] The purchase probability of a user for a target product can be calculated using the following formula:

[0144]

[0145] Among them, the coefficients (or weighted coefficients) α and β can be adjusted according to the scene and actual effect. The specific value range of α and β is [0,1], and α+β=1. It indicates the price matching degree of user u for product P (target product), the price of product P is p, Indicates that the brand the user purchased last time was v i , then the user will buy brand v next j The probability that brand v i With brand v j The vector similarity between j It is the brand that P belongs to.

[0146] S505: sorting the purchase probability (Prob) values ​​of the target products, taking the second preset number of target products with the highest purchase probability, and recommending them to the user.

[0147] The second preset quantity can be customized.

[0148] Based on the introduction of the above embodiments, the overall architecture of the product recommendation in the embodiment of the present invention is as follows: Figure 6 As shown, since each of the above embodiments has already introduced each part of the process in detail, it will not be repeated here.

[0149] The embodiment of the present invention mines category relationships based on purchase time intervals, mines brand purchase relationships based on different categories, and mines user price preferences. First, based on the user's purchase time interval, potential purchase categories are mined, then the purchase relationship between brands is mined according to the graph embedding method, and finally the user's price preference is mined according to the user's historical behavior. It should be noted that the graph embedding technology and solution of the embodiment of the present invention are not limited to the contents introduced in the above embodiment, and other methods can also be selected. For example, the graph embedding model can also specifically adopt the Node2vec-Scalable Feature Learning for Networks proposed by Stanford University.

[0150] The embodiments of the present invention take into account the conversion relationship between categories, enrich the diversity of recommendation results, and take the purchase time interval into consideration, enrich the research on category conversion, and fully explore the user's price preferences, so as to generate recommendation results more accurately. It can overcome the homogeneity defect of recommendation results in the prior art, capture the changes in user interests between different categories, and provide diverse, rich and accurate recommendation results according to the changes in user interests.

[0151] Figure 7 4 is a schematic diagram of main modules of a commodity recommendation device according to an embodiment of the present invention.

[0152] like Figure 7 As shown, a commodity recommendation device 700 according to an embodiment of the present invention mainly includes: a category prediction module 701 , a brand prediction module 702 , and a commodity recommendation module 703 .

[0153] The category prediction module 701 is used to determine the candidate category of the first user's next purchase of goods based on the category transfer data of the first user and each second user in the nearest neighbor set of the first user and the set recommendation time.

[0154] The brand prediction module 702 is used to determine the candidate brand of the first user's next purchase of the candidate category product by using the vector similarity between the first brand and each brand in the brand purchase network. The brand purchase network is constructed based on the purchase records of all users, and the first brand is the brand of the first user's most recent purchase in the past.

[0155] The product recommendation module 703 is used to calculate the purchase probability of the first user for each target product of the candidate brand based on the price matching degree of the first user for each target product of the candidate brand and the vector similarity between the first brand and the candidate brand in the brand purchase network, so as to recommend products of the candidate brand to the first user according to the purchase probability.

[0156] In one embodiment, the category transfer data of the user's past purchases of goods includes the category transfer pairs in the user's purchase category sequence, the frequency of occurrence of the category transfer pairs in the purchase category sequence, and the purchase time of the two categories in the category transfer pairs by the user. The users include a first user and a second user.

[0157] In one embodiment, the category prediction module 701 is specifically used to: for each second user, obtain a first set from the category transfer pairs in the purchase category sequence of the first user, obtain a second set from the category transfer pairs in the purchase category sequence of the second user, and calculate the similarity of the purchase category sequences of the first user and the second user according to the sum of the occurrence frequencies in the intersection of the first set and the second set, and the sum of the occurrence frequencies in the first set and the second set; for each second user, calculate the purchase time interval corresponding to the target category transfer pairs of the first category in the purchase category sequence of the second user, and the occurrence frequency of the target category transfer pairs in the purchase category sequence of the second user. The initial purchase time interval preference on the target category transfer pair, the first category is the category where the first brand is located, and the purchase time interval corresponding to the target category transfer pair is the difference between the purchase time of the second user in the two categories in the target category transfer pair; according to the similarity of the purchase category sequence between the first user and each second user, and the initial purchase time interval preference of each second user on the target category transfer pair, the final purchase time interval preference of the first user on the target category transfer pair is calculated; according to the purchase time of the first user in the first category, the set recommendation time, and the final purchase time interval preference of the first user on the target category transfer pair, the probability of each category that the first user will purchase the product next time is calculated, and the candidate category is selected from each category according to the probability.

[0158] In one embodiment, the brand prediction module 702 is specifically used to: learn the constructed brand purchase network through a preset model to obtain a brand vector representation with category personalization; and calculate the vector similarity between the first brand and each brand using the brand vector representation with category personalization.

[0159] In one embodiment, the product recommendation device 700 also includes a price matching calculation module, which is used to calculate the price matching degree of the first user for each target product of the candidate brand in the following manner: for each target product, the degree of membership of the target product's price to each price level is calculated based on the price of the target product and the lowest price and the highest price of products in the category to which the target product belongs; the user price sensitivity of the first user corresponding to each price level is calculated based on the degree of membership of the price of each target product to each price level and the first user's satisfaction with each target product; the price matching degree of the first user for each target product of the candidate brand is calculated based on the user price sensitivity of the first user corresponding to each price level and the degree of membership of the price of each target product to each price level.

[0160] In one embodiment, the product recommendation module 703 is used to: for each target product, sum and weight the price matching degree of the first user for the target product and the vector similarity between the first brand in the brand purchase network and the brand to which the target product belongs; and calculate the purchase probability of the first user for each target product based on the result of the weighted sum and the result of the sum.

[0161] In addition, the specific implementation content of the commodity recommendation device in the embodiment of the present invention has been described in detail in the above-mentioned commodity recommendation method, so the repeated content will not be described again here.

[0162] Figure 8 An exemplary system architecture 800 is shown to which the commodity recommendation method or commodity recommendation device according to the embodiment of the present invention can be applied.

[0163] like Figure 8 As shown, system architecture 800 may include terminal devices 801, 802, 803, a network 804, and a server 805. Network 804 is used to provide a medium for communication links between terminal devices 801, 802, 803 and server 805. Network 804 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0164] Users can use terminal devices 801, 802, and 803 to interact with server 805 through network 804 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 801, 802, and 803, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0165] The terminal devices 801 , 802 , and 803 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0166] The server 805 may be a server that provides various services, such as a backend management server (only an example) that provides support for shopping websites browsed by users using the terminal devices 801, 802, and 803. The backend management server may analyze and process the received data such as product information query requests, and feed back the processing results (such as target push information, product information - only an example) to the terminal device.

[0167] It should be noted that the commodity recommendation method provided in the embodiment of the present invention is generally executed by the server 805 , and accordingly, the commodity recommendation device is generally disposed in the server 805 .

[0168] It should be understood that Figure 8 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0169] Reference below Fig. 9 , which shows a schematic diagram of the structure of a computer system 900 of a terminal device or server suitable for implementing an embodiment of the present application. Fig. 9 The terminal device or server shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0170] like Fig. 9 As shown, the computer system 900 includes a central processing unit (CPU) 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage part 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the system 900 are also stored. The CPU 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0171] The following components are connected to the I / O interface 905: an input section 906 including a keyboard, a mouse, etc.; an output section 907 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, a modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the I / O interface 905 as needed. A removable medium 911, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 910 as needed, so that a computer program read therefrom is installed into the storage section 908 as needed.

[0172] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 909, and / or installed from the removable medium 911. When the computer program is executed by the central processing unit (CPU) 901, the above-mentioned functions defined in the system of the present application are executed.

[0173] It should be noted that the computer-readable medium shown in the present invention can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal can take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0174] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the above-mentioned module, program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0175] The modules involved in the embodiments of the present invention may be implemented in software or hardware. The modules described may also be provided in a processor, for example, may be described as: a processor including a category prediction module, a brand prediction module, and a product recommendation module. The names of these modules do not, in some cases, constitute limitations on the modules themselves. For example, the category prediction module may also be described as "a module for determining candidate categories for the next purchase of goods by the first user based on the category transfer data of the past purchases of goods by each second user in the first user and its nearest neighbor set and the set recommendation time."

[0176] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently and not be assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: determining the candidate category of the first user's next purchase of goods according to the category transfer data of the past purchase of goods by each second user in the first user and its nearest neighbor set and the set recommendation time; determining the candidate brand of the first user's next purchase of the goods of the candidate category by the first user using the vector similarity between the first brand and each brand in the brand purchase network, the brand purchase network is constructed based on all user purchase records, and the first brand is the brand of the first user's most recent purchase in the past; based on the price matching degree of each target product of the candidate brand by the first user, and the vector similarity between the first brand and the candidate brand in the brand purchase network, calculating the purchase probability of the first user for each target product, so as to recommend the candidate brand products to the first user according to the purchase probability.

[0177] According to the technical solution of the embodiment of the present invention, based on the category transfer data of the past purchases of the first user and each second user in the nearest neighbor set and the set recommendation time, the candidate category of the first user's next purchase is determined, and the candidate brand of the first user's next purchase of the candidate category is determined by using the vector similarity between the first brand and each brand in the brand purchase network; based on the price matching degree of the first user for each target product of the candidate brand and the vector similarity between the first brand and the candidate brand, the purchase probability of the first user for each target product is calculated to recommend products based on the purchase probability. It can capture the changes in user interests between different categories, take the purchase time interval into consideration, fully explore the user's price preference, and provide diverse, rich and accurate recommendation results based on the changes in user interests.

[0178] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A product recommendation method, characterized in that: include: Determine a candidate category for the next purchase of goods by the first user according to the category transfer data of the goods purchased in the past by the first user and each second user in the nearest neighbor set and the set recommendation time; The category transfer data of the user's past purchases of goods includes the category transfer pairs in the user's purchase category sequence, the frequency of occurrence of the category transfer pairs in the purchase category sequence, and the purchase time of the two categories in the category transfer pairs by the user, and the users include the first user and the second user; Determine a candidate brand of the product of the candidate category that the first user will purchase next time by using the vector similarity between the first brand and each brand in the brand purchase network, wherein the brand purchase network is constructed based on the purchase records of all users, and the first brand is the brand of the product that the first user purchased most recently in the past; Based on the price matching degree of each target product of the candidate brand for the first user and the vector similarity between the first brand and the candidate brand in the brand purchase network, the purchase probability of each target product by the first user is calculated, so as to recommend the product of the candidate brand to the first user according to the purchase probability; The method of determining the candidate category of the next purchase of the first user based on the category transfer data of the past purchases of the first user and each second user in the first user and its nearest neighbor set and the set recommendation time comprises: for each of the second users, obtaining a first set from the category transfer pairs in the purchase category sequence of the first user, obtaining a second set from the category transfer pairs in the purchase category sequence of the second user, and calculating the similarity of the purchase category sequence of the first user and the second user based on the sum of the occurrence frequencies in the intersection of the first set and the second set and the sum of the occurrence frequencies in the first set and the second set; for each of the second users, calculating the similarity of the purchase category sequence of the first user and the second user based on the purchase category of the second user; The purchase time interval corresponding to the target category transfer pair of the first category included in the purchase category sequence, and the occurrence frequency of the target category transfer pair in the purchase category sequence of the second user, calculate the initial purchase time interval preference of the second user on the target category transfer pair, the first category is the category where the first brand is located, and the purchase time interval corresponding to the target category transfer pair is the difference between the purchase time of the two categories in the target category transfer pair by the second user; according to the similarity of the purchase category sequences of the first user and each of the second users, and the initial purchase time interval preference of each of the second users on the target category transfer pair, calculate the final purchase time interval preference of the first user on the target category transfer pair; According to the purchase time of the first user in the first category, the set recommendation time, and the final purchase time interval preference of the first user on the target category transfer pair, the probability of each category for the first user to purchase goods next time is calculated, and the candidate category is selected from the various categories according to the probability.

2. The method according to claim 1, characterized in that The determining, by using the vector similarity between the first brand and each brand in the brand purchase network, a candidate brand for the first user to purchase the product of the candidate category next time includes: The constructed brand purchase network is learned through a preset model to obtain a brand vector representation with category personalization; The brand vector representation with category personalization is used to calculate the vector similarity between the first brand and each brand.

3. The method according to claim 1, characterized in that: The process further includes calculating the price matching degree of the first user for each target product of the candidate brand in the following manner: For each of the target products, the degree of membership of the target product's price to each price level is calculated according to the price of the target product and the lowest price and the highest price of products in the category to which the target product belongs; Calculating the user price sensitivity of the first user corresponding to each of the price levels according to the degree of membership of the price of each of the target products to each of the price levels and the first user's satisfaction with each of the target products; The price matching degree of the first user for each target product of the candidate brand is calculated according to the user price sensitivity of the first user corresponding to each price level and the degree of membership of the price of each target product to each price level.

4. The method according to claim 3, characterized in that The calculating the purchase probability of the first user for each target product of the candidate brand based on the price matching degree of the first user for each target product of the candidate brand and the vector similarity between the first brand and the candidate brand in the brand purchase network includes: For each of the target products, summing and weighting the price matching degree of the first user for the target product and the vector similarity between the first brand in the brand purchase network and the brand to which the target product belongs; The purchase probability of the first user for each of the target commodities is calculated based on the weighted sum result and the sum result.

5. A commodity recommendation device, characterized in that: include: A category prediction module, used to determine a candidate category for the next purchase of goods by the first user based on the category transfer data of the goods purchased in the past by the first user and each second user in the nearest neighbor set and the set recommendation time; The category transfer data of the user's past purchases of goods includes the category transfer pairs in the user's purchase category sequence, the frequency of occurrence of the category transfer pairs in the purchase category sequence, and the purchase time of the two categories in the category transfer pairs by the user, and the users include the first user and the second user; A brand prediction module, used to determine a candidate brand of the product of the candidate category that the first user will purchase next time by using the vector similarity between the first brand and each brand in the brand purchase network, wherein the brand purchase network is constructed based on the purchase records of all users, and the first brand is the brand of the product that the first user purchased most recently in the past; a product recommendation module, configured to calculate the purchase probability of the first user for each target product of the candidate brand based on the price matching degree of the first user for each target product of the candidate brand and the vector similarity between the first brand and the candidate brand in the brand purchase network, so as to recommend the product of the candidate brand to the first user according to the purchase probability; The category prediction module is specifically used to: for each second user, obtain a first set from the category transfer pairs in the purchase category sequence of the first user, obtain a second set from the category transfer pairs in the purchase category sequence of the second user, and calculate the similarity of the purchase category sequences of the first user and the second user according to the sum of the occurrence frequencies in the intersection of the first set and the second set, and the sum of the occurrence frequencies in the first set and the second set; For each second user, according to the purchase time interval corresponding to the target category transfer pair including the first category in the purchase category sequence of the second user, and the occurrence frequency of the target category transfer pair in the purchase category sequence of the second user, the initial purchase time interval preference of the second user on the target category transfer pair is calculated, the first category is the category where the first brand is located, and the purchase time interval corresponding to the target category transfer pair is the difference between the purchase time of the two categories in the target category transfer pair of the second user; according to the similarity of the purchase category sequence of the first user and each second user, and the initial purchase time interval preference of each second user on the target category transfer pair, the final purchase time interval preference of the first user on the target category transfer pair is calculated; According to the purchase time of the first user in the first category, the set recommendation time, and the final purchase time interval preference of the first user on the target category transfer pair, the probability of each category for the first user to purchase the product next time is calculated, and candidate categories are selected from each category according to the probability.

6. An electronic device, characterized in that: include: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method according to any one of claims 1 to 4.

7. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images